Results for “Prediction Accuracy” — Page 38

NASA NTRS2021

Raman Temperature and Density Measurements in Supersonic Jets

Prediction of flow-field properties in supersonic jets using computational fluid dynamics (CFD) code predictions has become routine; however, obtaining accurate solutions becomes more challenging when there is a significant temperature difference between the jet core and the ambient air and/or compressibility effects are significant. Benchmark sets of flow field property data are required in order to assess current CFD capabilities and develop better modeling approaches for these turbulent flow fields where accurate calculation of temperatures and turbulent heat flux is important. Particle Image Velocimetry (PIV), spontaneous rotational Raman scattering (SRS) spectroscopy and Background Oriented Schlieren (BOS) have been previously used to acquire measurements of the mean and root mean square (rms) velocities, the mean and rms gas temperatures and density gradients in subsonic jet flows and film cooling flows. In this work, the ability to measure density is added to the list of measurands available using the acquired Raman spectra. The suite of measurement techniques are now applied to supersonic jet flows. The computation of the local gas pressure in the potential core of an over-expanded jet is demonstrated using the Raman measured gas temperature and density. Additionally, a unique density feature in temperature matched, perfectly expanded jet flow shear layers identified using BOS was verified using the Raman measurement technique. These non-intrusive flow measurements are compared against RANS predictions of the supersonic jet flow properties as a means of assessing their prediction accuracy.

Raman Spectroscopy↗

NASA NTRSDate not supplied

Scaling Methodology for Buckling of Composite Conical Shells in Axial Compression

Conical shells are commonly used as structural components for launch vehicles. The axial compression experienced during launch is one of the sizing load cases, because it can lead to loss of stability. As Because experimentally testing these large full-scale structures is cumbersome and expensive, it is studied how reduced-scale shells can be designed such that their buckling behavior is representative of the full-scale scale shell behavior. An analytical, sequential scaling procedure methodology is developed based on the nondimensional governing equations for composite conical shells with a symmetric, balanced layup and negligible flexural anisotropy. The buckling behavior of the shells of different size is compared using linear and nonlinear finite element analyses, and good comparisons . Accurate results are obtained for the considered shells in terms of buckling load, displacement, and mode. The inclusion of a geometric imperfections affects reduces the prediction accuracy, but it does not to the extent that the methodology is no longer valid cause the methodology to fail.

Buckling↗

NASA NTRSDate not supplied

Experimental, Computational, Theoretical and Analytical Investigation of Flow Boiling in Reduced Gravity

Two-phase thermal management systems are far superior to their single-phase counterparts because of their ability to capitalize on the coolant’s both sensible and latent heats, thereby yielding orders of magnitude higher heat transfer coefficients and smaller system footprints. A vital knowledge necessary for their implementation in future space systems is performance in microgravity. Long-duration microgravity experiments are necessary to obtain reliable databases, which would then be used to build reliable predictive tools. To achieve this goal, investigators at the Purdue University Boiling and Two-Phase Flow Laboratory (PU-BTPFL) and the NASA Glenn Research Center (NASA-GRC) have been collaborating towards the development of the Flow Boiling and Condensation Experiment (FBCE) and eventual execution onboard the International Space Station (ISS). FBCE has now matured to a point where it is ready for transport to the ISS, where first tests will be conducted using the Flow Boiling Module (FBM). In preparation for the ISS tests, a series of pre-launch Mission Sequence Tests (MSTs) was performed at GRC in Earth gravity with FBM mounted in a vertical upflow orientation using n-perfluorohexane as working fluid. The pre-launch tests included variations of flow rate, surface heat flux, inlet conditions, and both single-sided and double-sided wall heating. This presentation will summarize experimental results from these tests as well as both analytic and theoretical tools for prediction of two-phase heat transfer coefficient and critical heat flux (CHF). Also discussed will be an assessment of predictive accuracy of these tools against the experimental data.

Mission Sequence tests↗

NASA NTRSDate not supplied

Design of a Lunar Plume-Surface Interaction Measurement System

Plume-surface interactions between a rocket plume and the lunar surface will be studied in-situ during two of NASA’s upcoming Commercial Lunar Payload Services Program missions. The payload, Stereo Cameras for Lunar Plume-Surface Studies (SCALPSS), will employ a multiple-camera photogrammetry system to obtain stereo images of the lunar regolith before, during, and after significant erosion and subsequent crater formation. The evaluation of measurement system capabilities and design process of the payload is informed by computational fluid dynamics predictions, accuracy modeling based on experimental data, camera simulation, and lander design, all of which are combined in the Virtual Diagnostic Interface. The second SCALPSS mission, traveling to the moon on Firefly Aerospace’s Blue Ghost lander in early 2023, aims to build on the design and complexity of the first payload, which is launching in early 2022 on Intuitive Machines’ Nova-C lander. The Blue Ghost SCALPSS mission will include two additional cameras and a total of three different lens focal lengths which will target specific points in the vehicle’s descent to obtain quantitative and accurate 3D reconstruction of the lunar surface both prior to and after crater formation.

photogrammetry↗

NASA NTRS2022

Flow Visualization, Heat Transfer, and Critical Heat Flux of Flow Boiling in Earth Gravity with Saturated Liquid-Vapor Mixture Inlet Conditions – In Preparation for Experiments Onboard the International Space Station

This study investigates flow boiling of n-Perfluorohexane with saturated two-phase mixture inlet in a rectangular channel of dimensions 114.6-mm heated length, 2.5-mm width, and 5 mm-height. The experiments were performed as part of the Mission Sequence Testing of the Flow Boiling and Condensation Experiment’s (FBCE) Flow Boiling Module (FBM) in the vertical upflow configuration in Earth gravity using the same experimental system that was launched to the International Space Station (ISS) in August 2021. The operating parameters varied are heating configuration (single- and double-sided), mass velocity (380 – 2400 kg/m2s), inlet quality (0.011 – 0.519), and inlet pressure (120 – 179 kPa). High-speed video photographs are presented to explain the two-phase flow patterns within the channel’s heated length. Flow patterns are constituted by low-density and high-density fronts moving along the channel, with the high-density fronts gradually reducing in length due to evaporation. Heat transfer results in terms of flow boiling curves, streamwise wall temperature profiles, streamwise heat transfer coefficient profiles, and average heat transfer coefficients are presented and trends discussed. CHF data from the present experiments are combined with prior databases to compile a consolidated FBCE-CHF database for saturated inlet to expand the ranges of operating conditions and include other flow orientations in Earth gravity. Experimental CHF trends are also discussed. The interfacial lift-off model shows a good CHF predictive accuracy evidenced by a mean absolute error of 11.97% for this consolidated database after constraining it to mass velocities greater than or equal to 500 kg/m2s. Finally, this study confirmed reliability of the upcoming ISS experiments for saturated inlet conditions and the collected Earth-gravity data will be compared to ISS microgravity data.

Flow boiling↗

NASA NTRS2022

NASA Engineering and Safety Center Technical Bulletin No. 22-04: Uncertainty Quantification of Reduced Order Structural Dynamic Models

Uncertainty quantification (UQ) provides statistical bounds on prediction accuracy based on finite element model (FEM) uncertainty. An alternate method for UQ, called the Hybrid Parametric Variation (HPV) combines a parametric variation of the Hurty/Craig-Bampton (HCB) fixed-interface (FI) modal frequencies with a nonparametric variation (NPV) method. This provides a UQ method that can be traced to test data, which can be updated as additional data and improved correlated models become available.

Uncertainty Quantification↗

NASA NTRSDate not supplied

Investigating Solar Radiation Pressure Modeling for Operations in Near Rectilinear Halo Orbit

NASA’s Gateway program will build a crew-tended station in an Earth-Moon Near Rectilinear Halo Orbit (NRHO). Deep space operations differ considerably from Low Earth Orbit (LEO) operations in the environmental modeling, orbit geometry, and propagation timespans in operations. The cislunar environment, as opposed to the LEO environment, lacks atmospheric drag and is simultaneously affected by the gravity of both the Earth and the Moon, and solar radiation pressure (SRP) has a significant effect. This paper investigates the impacts of various SRP models on prediction accuracy, attitude control accuracy, orbit determination performance, and computational burden.

Guidance↗

NASA NTRSDate not supplied

Mapping Process Model Results to Fracture Model Initial Conditions for Adhesive Bonding

A systematic approach is proposed for mapping adhesive bonding process outcomes to initial conditions for progressive damage analysis (PDA). Herein, two mapping procedures are developed: 1) direct mapping for residual stresses, strains, and deformations obtained from a process model; and 2) functional mapping for quantities such as bondline thickness, degree of cure, and porosity that are assumed to have a functional relationship with fracture toughness but are not discretely modeled in the PDA. The spatial distribution of functionally mapped quantities may be determined by process modeling or inspection. The functional maps between the quantity of interest and fracture toughness may be obtained from independent lower-scale numerical simulations or empirical test campaigns. The proposed mapping procedure links the adhesive bonding process to the structural performance. Therefore, by accounting for the effects of bondline nonuniformities in structural analysis, analytical predictive accuracy can be improved, and off-nominal conditions can be evaluated. The mapping procedures are demonstrated and verified with example problems.

Andrew C Bergan↗

NASA NTRSDate not supplied

Mapping Process Model Results to Fracture Model Initial Conditions for Adhesive Bonding

A systematic approach is proposed for mapping adhesive bonding process outcomes to initial conditions for progressive damage analysis (PDA). Herein, two mapping procedures are developed: 1) direct mapping for residual stresses, strains, and deformations obtained from a process model; and 2) functional mapping for quantities such as bondline thickness, degree of cure, and porosity that are assumed to have a functional relationship with fracture toughness but are not discretely modeled in the PDA. The spatial distribution of functionally mapped quantities may be determined by process modeling or inspection. The functional maps between the quantity of interest and fracture toughness may be obtained from independent lower-scale numerical simulations or empirical test campaigns. The proposed mapping procedure links the adhesive bonding process to the structural performance. Therefore, by accounting for the effects of bondline nonuniformities in structural analysis, analytical predictive accuracy can be improved, and off-nominal conditions can be evaluated. The mapping procedures are demonstrated and verified with example problems.

Andrew Bergan↗

NASA NTRSDate not supplied

On Generalizably Improving RANS Predictions of Flow Separation and Reattachment

This work presents progress in an ongoing data-driven turbulence modeling effort at NASA Langley Research Center to improve predictions for separated flows. Two augmentation strategies are presented – one to improve predictions for the point of reattachment after smooth-body separation, and another to improve predictions for flow separation due to prolonged adverse pressure gradients. The reattachment augmentation was inspired from lessons learned during attempts at using field inversion and machine learning (FIML); however, it is prescribed as a conventional analytic closed-form expression. Results show that, when introduced into the SST k-w model, it shortens the otherwise overpredicted separation bubble length for three cases characterized by different Reynolds numbers and flow geometries. The second augmentation, which addresses flow separation under prolonged adverse pressure gradients, is inferred via FIML and applied as a neural network within the Spalart-Allmaras model. It uses features that are chosen such that the presented results show an improved predictive accuracy for the training case without affecting other cases where the augmentation is not needed. Further testing is required to establish the generalizability of this augmentation. The rationale behind the chosen features that serve as inputs to both these augmentations is also presented for the benefit of the reader.

machine learning↗

NASA NTRSDate not supplied

Pretest Simulations of a Supersonic Mixing and Combustion Validation Experiment to Assess Sensitivities

The reliance on CFD simulations to develop, design, and optimize scramjet systems (or components) has become commonplace. This reliance inevitably hinges on the ability of the computational analyst to quantify the level of confidence in their computational results. Unfortunately, nearly all the measured data available for this assessment comes from antiquated experimental datasets, or from tests that focused on the extraction of scramjet system (or component) performance. The objective of a CFD validation experiment is to quantify the predictive accuracy of one or more of the CFD physics submodels, implying that other uncertainties related to replicating the facility flow environment (e.g., knowledge of boundary conditions) must be minimized to the extent possible. This inevitably places stringent requirements on the quality and quantity of measurements taken to accurately specify inflow, outflow, and surface conditions for the CFD simulations; in addition to the measurements taken for the validation of physics submodels. This places additional demands on the experimental process above and beyond those for test article performance assessment. A recent high speed code credibility workshop series sponsored by AFRL identified a gap in existing validation data for fundamental assessments of turbulent mixing and combustion CFD closure models at scramjet engine relevant conditions. To address this gap, engineers at AFRL have designed a coaxial jet flame configuration that will be tested at two facilities (Research Cell 19 at the Air Force Research Lab, and at Purdue University). The effort described here documents pretest simulations of this validation experiment with the goal of fleshing out the extent of the facility flowpath that must be included to adequately reproduce the facility test section flow environment. The findings indicate that the flow around the support structure for the fuel injection centerbody upstream of the facility nozzle generates disturbances that persist throughout the nozzle expansion process; corrupting the azimuthal symmetry that was desired in the fuel/air mixing region of the test section. Simulations without this support structure maintained a high degree of azimuthal symmetry up until the fuel injection plane. However, even in this scenario the azimuthal symmetry was not maintained once the centerbody boundary layer transitioned to a wake flow downstream of the fuel injection plane.

CFD↗

NASA NTRSDate not supplied

Pretest Simulations of a Supersonic Mixing and Combustion Validation Experiment to Assess Sensitivities

The reliance on CFD simulations to develop, design, and optimize scramjet systems (or components) has become commonplace. This reliance inevitably hinges on the ability of the computational analyst to quantify the level of confidence in their computational results. Unfortunately, nearly all the measured data available for this assessment comes from antiquated experimental datasets, or from tests that focused on the extraction of scramjet system (or component) performance. The objective of a CFD validation experiment is to quantify the predictive accuracy of one or more of the CFD physics submodels, implying that other uncertainties related to replicating the facility flow environment (e.g., knowledge of boundary conditions) must be minimized to the extent possible. This inevitably places stringent requirements on the quality and quantity of measurements taken to accurately specify inflow, outflow, and surface conditions for the CFD simulations; in addition to the measurements taken for the validation of physics submodels. This places additional demands on the experimental process above and beyond those for test article performance assessment. A recent high speed code credibility workshop series sponsored by AFRL identified a gap in existing validation data for fundamental assessments of turbulent mixing and combustion CFD closure models at scramjet engine relevant conditions. To address this gap, engineers at AFRL have designed a coaxial jet flame configuration that will be tested at two facilities (Research Cell 19 at the Air Force Research Lab, and at Purdue University). The effort described here documents pretest simulations of this validation experiment with the goal of fleshing out the extent of the facility flowpath that must be included to adequately reproduce the facility test section flow environment. The findings indicate that the flow around the support structure for the fuel injection centerbody upstream of the facility nozzle generates disturbances that persist throughout the nozzle expansion process; corrupting the azimuthal symmetry that was desired in the fuel/air mixing region of the test section. Simulations without this support structure maintained a high degree of azimuthal symmetry up until the fuel injection plane. However, even in this scenario the azimuthal symmetry was not maintained once the centerbody boundary layer transitioned to a wake flow downstream of the fuel injection plane.

CFD↗

NASA NTRSDate not supplied

James Webb Space Telescope Navigation Optimization Challenges

The James Webb Space Telescope (JWST) is a NASA flagship mission launched on December 25, 2021. The early orbit phase was highlighted by three midcourse correction burns that were performed to maneuver the vehicle into a libration orbit at the second Sun-Earth-Moon (SEM) libration point (L2). During the coast to L2, the vehicle’s primary observatory and sunshield were deployed. Once the tennis-court-sized sunshield was unfurled and tensioned into place, the large area exposed to solar radiation pressure (SRP) dictated that the SRP force model would need to account for the vehicle’s geometry, reflective properties, and orientation relative to the Sun. Following the insertion of JWST into its L2 libration orbit on January 24, 2022, the commissioning phase for the vehicle’s observatory commenced along with the first cycle of routine station-keeping maneuvers. The NASA Goddard Space Flight Center’s Flight Dynamics Facility (FDF) provides navigation services for the JWST mission. The FDF serves as the prime or backup navigation operations center for more than 30 active missions spanning a wide array of flight regimes performing orbit determination, maneuver planning, trajectory optimization, and tracking data evaluation. Definitive orbit determination for JWST is performed by the FDF using an Extended Kalman Filter (EKF) to estimate the vehicle’s trajectory and reflective properties using Deep Space Network (DSN) TRK-2-34 tracking data and spacecraft attitude telemetry. For the purposes of orbit prediction and station-keeping maneuver targeting, the sensitivity of the SRP force to the vehicle’s orientation requires that predictive attitude information must be incorporated to generate accurate orbit predictions and maneuver plans. Predictive attitude plans are provided to the FDF by the JWST Spacecraft Operations Center (SOC). Short-term (28 day) orbit predictions are propagated using a Short Range Attitude Plan (SRAP) to model the vehicle’s future attitude states for up to a week. SRAP attitude data is highly reliable, as it reflects finalized planned attitude states that the vehicle will be commanded to attain during the ensuing week. A Long Range Attitude Plan (LRAP) can be implemented to model predicted attitude states beyond one week, but these predictions are tentative and subject to revision due to changes in science operations plans. Instead, a conservative approach of applying a Sun-Pointing Neutral (SPN) attitude configuration is utilized for long-term (2 year) orbit predictions where reliable planned attitude data is not available. SPN attitude mode aligns the net SRP force along the JWST-to-Sun vector where it becomes independent of the vehicle’s angle of rotation about this vector, defined as the Sun yaw. Given the dynamic instability of the libration orbit, station-keeping thrust must be applied in either the sunward or anti-sunward direction to balance the resulting orbit. The attitude constraints of the vehicle also impose limits on the available pointing directions for station-keeping thrusters. The thrusters cannot be aligned with the optimal pointing direction for sunward maneuvers, rendering sunward maneuvers to be significantly less fuel efficient than anti-sunward maneuvers. Consequently, station-keeping maneuvers must be targeted to balance the orbit while ensuring that the next maneuver will also be performed in the anti-sunward direction to optimize propellant usage. When targeting a station-keeping maneuver, the attitude and SRP modeling configuration which is applied to the predicted post-maneuver trajectory will dictate the direction of the subsequent station-keeping maneuver, assuming nominal propulsion system performance. If the predicted post-maneuver attitude states result in an SRP model which under-predicts the cumulative SRP impact on the post-maneuver orbit, a targeted anti-sunward maneuver will be larger than necessary and the next maneuver will need to be executed sunward in order to compensate. In contrast, an anti-sunward maneuver which is targeted using an SRP model which over-predicts the cumulative SRP impact will achieve station-keeping while ensuring that the next maneuver will likewise be performed anti-sunward. For this reason, station-keeping maneuvers are targeted while applying the SPN attitude mode, as this mode entails the largest possible SRP area cross-section and therefore a larger modeled cumulative SRP impact. This paper documents the NASA Goddard Space Flight Center’s FDF support for JWST on-orbit operations and the analysis projects undertaken utilizing the experiences and data accumulated throughout the first full year of routine science operations. The results of these analysis efforts have been used to implement improvements to orbit prediction accuracy and maneuver efficiency which have the potential to prolong the lifespan of JWST to continue to conduct ground-breaking infrared astronomy for decades to come.

Flight Dynamics↗

NASA NTRSDate not supplied

Mastering the Cryogenic Frontier: Predicting Sloshing in Cryogenic Propellant Tanks

In recent years, NASA has undertaken the development and validation of computation fluid dynamic (CFD) models for predicting cryogenic two-phase sloshing within propellant tanks. These models have been rigorously anchored against cryogenic ground tests that induce sloshing, as well as limited microgravity tests. The primary focus of these models is to accurately predict the two-phase condensation and, at times, evaporation within the cryogenic propellant. This approach is important for estimating the required autogenous and/or helium pressurant for completing propulsion system burns or ensuring adequate pressure for in-space propellant transfers. Unlike in hypergolic propulsion systems, where the propellant is stored at room temperature, cryogenic propellant is highly volatile. Therefore, ensuring the accurate estimation of pressurant necessitates the precise modeling of two-phase interfacial heat and mass transfer processes. This paper aims to provide an overview of NASA’s continuing efforts in developing and validating these CFD models, with particular emphasis on benchmark cases with sloshing that enhances phase change, tracing their historical development, and demonstrating their predictive accuracy when compared to test results.

low-gravity fluid modeling↗

NASA NTRSDate not supplied

Low-Speed Testing of the 5.2% Semispan CRM-HL in the DNW-NWB Wind Tunnel

In this paper the wind tunnel test of the 5.2% CRM-HL in the DNW-NWB low speed wind tunnel is described. The tunnel test campaign is settled in the DLR project ADaMant, in which it is of interest to generate qualitative data for CFD validation to improve the predictive accuracy, efficiency, and usability of physical modeling in CFD codes at the borders of the flight envelope. Different configurations of the CRM-HL were tested in a closed as well as in an open test section of the wind tunnel. Several measurement techniques were used to generate validation data including forces and moments, pressures, flow visualization with tufts and oil painting, and infrared thermography for boundary layer transition detection. The evaluation of the results focuses on these techniques in the closed test section. The test comprises Mach numbers from 0.15 to 0.23 and Reynolds numbers from 1.2x10 6 to 1.2x10 6 . Some results are also compared to data from the NASA Langley Research Center 14x22 Foot wind tunnel to assess the repeatability of measurements of the same model in different wind tunnels.

DLR↗

NASA NTRS2024

A High-Performance Computing Predictive GNSS Performance Monitor for Autonomous Air Vehicles in Urban Environments

This report offers analysis and design insights for leveraging High-Performance Computing (HPC) to predict line-of-sight (LOS) Global Navigation Satellite System (GNSS) availability in a city. This work is motivated by the emerging fields of Advanced and Urban Air Mobility (AAM/UAM), where regulatory authorities are seeking city-scale, meter-resolution risk forecasting in order to safely integrate new flight missions with existing urban life and infrastructure. This work addresses the technical challenge of efficiently computing urban GNSS satellite visibility to predict GNSS performance metrics under these requirements. We present a new HPC-optimized shadow casting algorithm variant as a ray-based approach to forecasting satellite visibility. We apply this algorithm variant in a software-defined prognostic service which generates a GNSS navigation risk-correlated map as a path planning-style potential field. We detail dominant computational burdens, viable simplifying assumptions, and different algorithmic implementations, intending to demonstrate a baseline of computation time needed by each stage in such a service. We conclude by analyzing the prototype service’s prediction accuracy compared to receiver data from Corpus Christi, Texas. This informs design trade-offs along the dimensions of hardware, computation time, and tolerable forecasting error (including proportions of false positives and false negatives).

GNSS↗

NASA NTRSDate not supplied

A Novel Framework for Multi-Path Data Fusion in Earth Observation and New Observing Strategies: Applications to Predicting Forest Canopy Height

Exponential growth of data from Earth Observation (EO) assets has necessitated the development of sophisticated methods for data interpretation and management. NASA’s New Observing Strategy (NOS) approach aims to coordinate operations among complex heterogenous systems of constellations, requiring advanced Artificial Intelligence and Machine Learning (AI/ML) techniques. Despite significant advancements in AI/ML across various domains, the EO and machine learning for satellite (SatML) fields remain fragmented, often relying on adapted techniques rather than domain-specific solutions. We present a novel end-to-end data fusion framework tailored specifically for EO and SatML, addressing this gap by facilitating rapid development of AI/ML applications. This framework, called, Multimodal Earth Observation Workflow for Machine Learning (MEOW-ML), sup- ports the entire AI/ML lifecycle, from dataset manipulation, to model training, evaluation, and logging, and is designed to expedite the development of next-generation NOS deployments and SOTA in EO. We apply our framework to predict canopy height model (CHM) derived from lidar data. We integrate multiple data modalities through a hierarchical, multi-path model architecture, effectively identifying and leveraging the unique strengths of each data source to enhance predictive accuracy. Our experiments demonstrate that the multi-path architecture outperforms traditional single-path models and provides significant advantages in both accuracy and computational efficiency.

Mark Moussa↗

NASA NTRS2024

Evaluation of Machine Learning and Deep Learning Algorithms for Fire Prediction in Southeast Asia

Vegetation fires are prevalent in South/Southeast Asian countries, making fire prediction crucial due to their potential environmental, economic, and social impacts. Accurate predictions of fires facilitate timely interventions, helping to mitigate uncontrolled fires that can lead to biodiversity loss and air quality issues. In this study, we utilize VIIRS satellite-derived fire data alongside six machine learning and deep learning models—Simple Persistence, Multi-Layer Perceptron (MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), CNN-LSTM, and ConvLSTM—to determine the most effective fire prediction model, using Root Mean Square Error (RMSE) as the metric. Our results indicate that the CNN model is the most reliable in regions with spatial dependencies, such as Brunei, Indonesia, Malaysia, the Philippines, Timor-Leste, and Thailand. Conversely, the ConvLSTM model excels in countries with complex spatiotemporal dynamics like Laos, Myanmar, and Vietnam. The CNN-LSTM hybrid model also performed well in Cambodia, suggesting a need for a balanced approach in areas requiring both spatial and temporal feature extraction. Furthermore, simpler models like Persistence and MLP showed limitations in capturing dynamic patterns and temporal dependencies. Our findings highlight the importance of evaluating models before implementing any decision support systems (DSS) in fire management. By tailoring models to specific regional fire data, we can enhance prediction accuracy and responsiveness, ultimately improving fire risk management in Southeast Asia and beyond.

Deep learning↗